OUP user menu

Left atrial remodelling in patients with myocardial infarction complicated by heart failure, left ventricular dysfunction, or both: the VALIANT Echo Study

Alessandra Meris , Maria Amigoni , Hajime Uno , Jens Jakob Thune , Anil Verma , Lars Køber , Mikhail Bourgoun , John J. McMurray , Eric J. Velazquez , Aldo P. Maggioni , Jalal Ghali , J. Malcolm O. Arnold , Steven Zelenkofske , Marc A. Pfeffer , Scott D. Solomon
DOI: http://dx.doi.org/10.1093/eurheartj/ehn499 56-65 First published online: 11 November 2008


Aims To assess the relationship between left atrial (LA) size and outcome after high-risk myocardial infarction (MI) and to study dynamic changes in LA size during long-term follow-up.

Methods and results The VALIANT Echocardiography study prospectively enrolled 610 patients with left ventricular (LV) dysfunction, heart failure (HF), or both following MI. We assessed LA volume indexed to body surface area (LAVi) at baseline, 1 month, and 20 months after MI. Baseline LAVi was an independent predictor of all-cause death or HF hospitalization (P = 0.004). In patients who survived to 20 months, LAVi increased a mean of 3.00 ± 7.08 mL/m2 from baseline. Hypertension, lower estimated glomerular filtration rate, and LV mass were the only baseline independent predictors of LA remodelling. Changes in LA size were related to worsening in MR and increasing in LV volumes. LA enlargement during the first month was significantly greater in patients who subsequently died or were hospitalized for HF than in patients without events.

Conclusion Baseline LA size is an independent predictor of death or HF hospitalization following high-risk MI. Moreover, LA remodelling during the first month after infarction is associated with adverse outcome.

  • Echocardiography
  • Left atrial remodelling
  • Myocardial infarction
  • Prognosis


Left atrial (LA) size is an important predictor of prognosis both in the general population and in patients with heart disease,1 including left ventricular (LV) dysfunction,2 mitral regurgitation (MR),3 or atrial fibrillation (AF).4 LA volume is a powerful and independent predictor of mortality after myocardial infarction (MI)5 with smaller LA volumes associated with good prognosis even in patients with depressed systolic function.6 In patients with acute MI, LA volume may serve as a surrogate measure of chronic diastolic function and ventricular filling pressure, less affected by acute haemodynamic changes than transmitral Doppler measures.710

We studied patients enrolled in the VALsartan In Acute myocardial iNfarcTion (VALIANT) Echo study to test the hypothesis that LA volume is a predictor of both mortality and cardiovascular morbidity after high-risk MI and to assess the severity and timing of LA remodelling during early and long-term follow-up after MI.


Study population

The VALIANT study was designed to test the efficacy and safety of long-term treatment with valsartan, captopril, and their combination after acute MI in 14 703 patients with LV dysfunction (LV ejection fraction ≤35% on echocardiography or contrast angiogram, or ≤40% on radionuclide ventriculography), clinical signs of heart failure (HF), or both. Inclusion and exclusion criteria, and the details of patient characteristics have been previously described.11

The VALIANT Echo study was designed prospectively to test the hypothesis that valsartan, either alone or in combination with captopril, could attenuate progressive LV enlargement or improve LV function to a greater extent than captopril alone.12 Ninety-four of all the 931 clinical centres participating in VALIANT accepted the invitation to enrol patients in the VALIANT Echo study, and 610 patients from VALIANT were enrolled in the Echo study. Echocardiographic examinations were performed at a mean of 4.9 days after MI (before randomization), at 1 month, and at 20 months. The median follow-up time was 20.4 months (interquartile range: 17.9–23.4 months). Except for minor differences, the VALIANT Echo study population was similar to the overall VALIANT cohort. Patients signed informed consent for inclusion in the VALIANT Echo study, and the protocol was approved by the appropriate institutional review boards.

Echocardiographic analysis

All echocardiographic studies were analysed in the core laboratory at Brigham and Women’s Hospital. Studies were reviewed for quality and five patients were excluded before analysis because of extremely poor study quality. Therefore, the whole echocardiographic cohort was composed of 605 patients; another 39 patients had insufficient echocardiographic images for endocardial border definition of the left atrium and were excluded from this study. Echocardiographic analyses were performed in 566 patients at baseline, in 507 at 1 month, and in 379 patients at 20 months. Of the initial cohort, 59 did not have 1 month echocardiograms (including 18 who died during the first month) and 187 did not have 20 months echocardiograms (including 99 who died before the final echocardiographic control). An additional five patients died after the 20 months echocardiograms.

Echocardiograms from videotape were digitized and analyses were performed with the use of offline quantitative analysis software. LV endocardial borders were manually traced at end-diastole and end-systole in apical four- and two-chamber views. LV volumes were derived according to the modified biplane Simpson’s rule and LV ejection fraction (LV-EF) was calculated in the standard fashion from ventricular volumes. LV mass was calculated from LV end-diastolic diameter and septal and posterior wall thickness, following the current guidelines.13 LA volume was assessed by the biplane area–length method from apical four- and two-chamber views at end-systole from the frame preceding mitral valve opening. LA volume index (LAVi) was calculated as LA volume to body surface area ratio (mL/m2). MR was categorized by a semiquantitative method mapping systolic jet expansion in the LA in four- and two-chamber views. MR was considered mild when regurgitant jet area occupied >5 and <20% of the LA area, moderate when regurgitant jet area occupied >20 and <40%, and severe when regurgitant jet area occupied more than 40% of LA area.14,15 The presence of an eccentric jet raised the grade of MR by 1 degree.16 Mitral flow velocity was assessed by pulsed wave Doppler from the apical four-chamber view with measurement of the E and A waves (in sinus rhythm) velocity, their ratio, and the deceleration time of the E wave. Right ventricular function was quantified by right ventricular fractional area change (RV-FAC) as the ratio between RV end-diastolic area minus RV end-systolic area and RV end-diastolic area. All the echocardiographic measurements were repeated in three separate cardiac cycles.

Statistical methods

LAVi measurements were performed by two different readers: the first reader measured LAVi at baseline and the second reader measured LAVi at one month and at 20 months. The interobserver reproducibility was tested in 30 randomly selected patients and the limits of agreements were between −1.83 and 2.22 mL/m2 by Bland & Altman method.17 The 95% confidence interval for the lower limit of agreement was −2.50 to −1.17, and that for the upper limit of agreement was 1.56 to 2.88. To assess intraobserver reproducibility for LAVi, the second reader performed repeated measurements within few days in 30 randomly selected patients and the limits of agreements were between −2.06 and 1.53 mL/m2 by Bland & Altman method. The 95% confidence interval for the lower limit of agreement was −2.64 to −1.47, and that for the upper limit of agreement was 0.95 to 2.11. The coefficients of variability for intra- and interobsever reproducibility were 5.4 and 3.5%, respectively.

Baseline continuous data are expressed as mean ± standard deviation. Patients were divided according to LAVi at baseline into three categories (≤26, 26–32, and >32 mL/m2). The cut-offs of 26 and 32 mL/m2 were chosen at one and two standard-deviations from the normal LA size and have been previously validated in relation to clinically relevant endpoints.5,6,18 LAVi of greater than 26 mL/m2 identifies patients with an enlarged LA. Increased LAVi greater than 32 mL/m2 is a known marker for poor prognosis after MI. LAVi changes during follow-up were presented in quartiles. The relationship between the variable of interest (baseline LAVi or LA remodelling) and other baseline characteristics was assessed by a regression analysis for baseline continuous variables and by a t-test for baseline binary variables.

LA remodelling was assessed by looking at the change in LAVi from baseline to 1 and 20 months. To assess the baseline predictors of LA remodelling, we performed a stepwise multivariable linear regression analysis applying a backward selection with a P-value of 0.2 and using 27 covariables that are already known to be predictive of outcome and LA dimensions from previous VALIANT analyses and from the literature;12,19,20 these covariates included age, gender, estimated glomerular filtration rate (eGFR), Killip class, history of MI, history of HF, percutaneous coronary intervention after index MI, AF after index MI, history of chronic pulmonary disease, new left bundle-branch block, history of angina, history of diabetes, history of hypertension, systolic and diastolic blood pressure, primary percutaneous coronary intervention, thrombolysis, anterior MI, Q-wave and non-Q-wave MI, LV-EF, LV end-diastolic volume (LV-EDV), LV end-systolic volume (LV-ESV), MR degree, restrictive diastolic pattern, RV-FAC, LV mass. Change in LAVi was compared with changes in other echocardiographic measurements using Pearson correlation.

The following clinical endpoints were assessed during the follow-up: all-cause mortality, cardiovascular mortality, hospitalization for HF, and a composite end-point of all-cause mortality and hospitalization for HF.11,12,21 The event rate was calculated by the Kaplan–Meier method for all-cause mortality and the composite end-point; it was derived from the cumulative incidence function for cardiovascular death and hospitalization for HF in order to account for competing risks.

The influence of baseline LAVi and LA remodelling in the first month on the risk of clinical outcomes was evaluated by Cox proportional hazards models. For a multivariable model, we included variables proved to be powerful baseline clinical predictors of mortality from the overall VALIANT study (age, eGFR, Killip class, history of MI, history of HF, percutaneous coronary intervention after index MI, AF after index MI, history of diabetes, history of chronic pulmonary disease, new left bundle-branch block, and history of angina)11,12 and echocardiographic parameters (LV-EF, severe MR, and RV-FAC), which were found to have a prognostic value in prior studies.12,19,20 In addition to the complete case analysis, the multiple imputation method was used to assess the effect of missing values.

A cut-off value for LA remodelling at 1 month was selected to maximize the prediction accuracy of the composite endpoint at 12 month. The variability of the prediction accuracy associated with the selected cut-off was evaluated by utilizing a bootstrap method with 1000 bootstrap samples. The other accuracy measures were presented with 10-fold cross-validation estimates and corresponding 95% confidence intervals. The inverse probability of censoring weighting was used to account for observations censored before 12 months.

All P-values were two-sided; P-value <0.05 was considered statistically significant. All statistical analyses were performed using STATA software, version 8 (Stata Corp., College Station, TX) and R version 2.7.0.


Baseline characteristics

The mean LAVi at baseline was 24.85 ± 8.29 mL/m2, was greater than 26 mL/m2 in 181 patients (32%), and greater than 32 mL/m2 in 85 patients (15%). Baseline characteristics according to LAVi are shown in Table 1. A larger LAVi was associated with older age, female sex, worse Killip class, lower eGFR, new left bundle-branch block, and non-Q-wave infarction. Patients with larger LAVi were more likely to have had a history of hypertension, diabetes, previous MI, and HF, and were more likely to have developed AF post-MI. Acute reperfusion therapy, such as thrombolysis or primary percutaneous coronary angioplasty, was more common in the group of patients with normal LAVi.

View this table:
Table 1

Baseline clinical and echocardiographic characteristics according to baseline left atrial volume index

LAVi was associated with worse MR at baseline, larger LV volumes, reduced LV-EF, higher LV mass, and worse RV function. Patients with larger left atria were more likely to have a diastolic restrictive pattern, as defined by deceleration time ≤140 ms and E wave to A wave ratio ≥1.5.

Clinical outcomes

In the Echo cohort, 104 patients (17%) died and 164 patients (27%) were hospitalized for HF and/or died. In a multivariable analysis, an LAVi >32 mL/m2 was found to be independently associated with the composite outcome of death or hospitalization for HF [adjusted HR (95% CI) 2.35 (1.28–4.31), P = 0.006] and with overall mortality (adjusted HR (95% CI) 2.36 (1.08–5.18), P = 0.031] (Table 2 and Figure 1). The hazard ratio for death or hospitalization for HF associated with each mL/m2 increase in LAVi beyond 26 mL/m2 was 1.07 [1.05–1.09] (Figure 2).

Figure 1

Kaplan–Meier estimates of the rate of death or hospitalization for heart failure according to baseline left atrial volume index. HF, heart failure; LAVi, left atrial volume index.

Figure 2

Hazard ratio for death or hospitalization for heart failure associated with each mL/m2 increase in left atrial volume index. CI, confidence intervals; HF, heart failure; HR, hazard ratio; LAVi, left atrial volume index.

View this table:
Table 2

Baseline left atrial volume index and risk of adverse eventsa

Left atrial remodelling

In the 379 patients with completed echocardiographic follow-up, the LAVi significantly increased from 23.1 ± 7.5 to 24.8 ± 7.9 mL/m2 at 1 month (P < 0.001) and to 26.1 ± 9.1 mL/m2 at 20 months (P < 0.001). The overall increase in LAVi was 3.00 ± 7.08 mL/m2 by 20 months. Greater increases in LA volume from baseline to 20 months were observed in patients with a history of hypertension and diabetes, lower eGFR, and higher LV mass. LA remodelling was not significantly related to age, baseline Killip class, AF post-MI, ventricular volumes, LV-EF, and MR (Table 3). By stepwise multivariable linear regression analysis, eGFR, history of hypertension, systolic blood pressure, and LV mass were the only baseline independent predictors of long-term LA enlargement after MI (P = 0.037, P = 0.026, P = 0.028, and P = 0.004, respectively) and they had additive and incremental value in predicting LA remodelling; there was no significant interaction between history of hypertension and eGFR with respect to changes in LA volume (P = 0.161). Patients in the worst quartile of LA remodelling had hypertension (P < 0.001) and lower eGFR (P = 0.001) compared with those in the best quartile. LA remodelling was significantly and powerfully related to increasing in MR jet area (r = 0.61, P < 0.001) during the entire follow-up period. Moreover, LA remodelling was significantly related to changes in LV end-diastolic (r = 0.17, P = 0.001) and end-systolic (r = 0.20, P < 0.001) volumes, and systolic sphericity index (r = 0.11, P = 0.032) over 20 months.

View this table:
Table 3

Baseline clinical and echocardiographic characteristics according to quartile of changes in left atrial volume index (mL/m2) during 20 months after myocardial infarction

Left atrial remodelling and outcome

Early LAVi remodelling during the first month post-MI was a significant predictor of all-cause mortality, HF hospitalization, or the composite outcome of all-cause mortality or hospitalization for HF, even after adjustment for baseline LAVi and covariates (Table 4). The cut-off value of LA remodelling at 1 month that minimized the prediction error for the composite endpoint of all-cause death or HF hospitalization within 12 months was 9 mL/m2. Figure 3 shows the Kaplan–Meier estimates of the rate of death or hospitalization for HF according to this classification. The prediction accuracy with this cut-off was 0.82. When this cut-off value was applied to 1000 of bootstrap samples with the same size, the range of the prediction accuracy was 0.78–0.85.

Figure 3

Kaplan–Meier estimates of the rate of death or hospitalization for heart failure, according to change in left atrial volume index from baseline to 1 month. HF, heart failure; LAVi, left atrial volume index. The analysis was restricted to patients who survived to more than 1 month.

View this table:
Table 4

Risk of subsequent adverse events for each mL/m2 increase in left atrial volume index at 1 month compared with baseline size

Cross-validation estimates and corresponding 95% confidence interval for a sensitivity, a specificity, a positive predictive value, and a negative predictive value were 0.11 (0.04–0.23), 0.98 (0.73–0.99), 0.59 (0.30–0.78), and 0.83 (0.78–0.86), respectively.


We have shown that baseline LA size is an independent predictor of death or hospitalization for HF in patients with high-risk MI. Moreover, LA remodelling begins early after MI, and is influenced by hypertension, renal impairment, and LV mass. Early LA remodelling is predictive of mortality and cardiovascular morbidity.

LA size is considered to be an expression of the diastolic burden, and increased LA volume usually reflects elevated ventricular filling pressure. During ventricular diastole, the left atrium is directly exposed to LV pressure through the open mitral valve. As an adaptation to the decreased ventricular compliance following MI, LA pressure rises, increasing LA wall tension and stretching the atrial myocardium; LA volume reflects the duration and severity of increased LA pressure.810 In the setting of an acute MI, patients with higher chronic LV filling pressure and worse previous diastolic dysfunction have lower haemodynamic ‘cardiac reserve’ that helps to withstand an acute decrease in myocardial contractility.6

Early after MI, LA size has been shown to provide prognostic information incremental to clinical data and standard echocardiographic predictors of outcome.5,6 The present study confirms and extends these conclusions to a population of patients with high-risk MI and shows that LA size is a predictor not only of mortality but also of cardiovascular morbidity. The increased risk associated with greater LA size appears to be continuous, and even patients with mild increases in LA size are at increased risk. We observed that in the acute phase post-MI LA size was a better prognostic predictor of outcome than transmitral Doppler indices. Indeed, Doppler indices may be quite sensitive to acute changes in the loading conditions secondary to HF and/or to drugs.22,23 On the contrary, LA volume is likely to be less affected by acute haemodynamic changes and may represent a more stable indicator of the duration and severity of diastolic function and filling pressure over time.8 Likewise, deceleration time, which is a potent predictor of outcome when in the frankly restrictive range,24 is less useful when >140 ms.

After an acute MI, ventricular structural and functional alterations can lead to ventricular relaxation abnormalities resulting in worsening of diastolic dysfunction and LA remodelling. Our data suggest that this process begins early after MI, even during the first month after the acute event and continues subsequently. Moreover, the rate of LA remodelling does not appear to occur in a linear fashion after MI, but appears to be greatest during the early phase. These data suggest that post-MI LA remodelling is, in part, a subacute maladaptive response to an acute event.

In our population of patients with high-risk MI, LA size was associated with hypertension, impaired renal function, diabetes, and LV mass, all important predictors of LA enlargement and outcome in this population.2527 All these factors are already known to increase LV filling pressure and therefore influence LA volume. However, we did not observe a significant association between LA remodelling and MR degree at baseline, which may simply reflect the dynamic nature of MR in the early post-MI period. Indeed, over time progressive LA remodelling was related to long-term MR. Additionally, patients with the greatest degree of MR are most likely to die prior to the follow-up assessment of LA size; this survivor bias may also have attenuated the relationship between baseline MR and LA remodelling.

Some limitations of the present study should be noted. As patients were not followed after the 20 months echocardiographic examination, we cannot assess the predictive value of late LA remodelling on subsequent mortality and morbidity. Our analysis of long-term atrial remodelling is subject to survivor bias, as many patients did not survive to have a 1 month or a 20 months echocardiogram. Because more sophisticated measures of diastolic function, such as Doppler Tissue Imaging, were not available in VALIANT, we cannot relate changes in atrial size to robust measures of diastolic function. In this study, we measured MR degree by a semi-quantitative method. Rigorous quantitative methods such as the PISA method are more challenging to obtain in a clinical trial, and were not utilized in the VALIANT Echo study. Nevertheless, semi-quantitative methods such as the ones used in this study have been shown to correlate well with more quantitative methods.20,2830 Finally, although mitral valve surgery was not a pre-specified outcome in VALIANT and data on this was not collected, only 10 patients in the Echo cohort underwent subsequent cardiac surgery, suggesting that the number of patients who might have had mitral valve surgery is likely to be very low.


After high-risk MI, LA size is an independent predictor of both mortality and cardiovascular morbidity. Moreover, changes in LA size in the month following MI predict subsequent adverse outcomes. These data suggest that assessment of LAVi and changes in LAVi following infarction represent important additions to the post-MI echocardiographic evaluation.


The VALIANT trial was funded by a grant from Novartis Pharmaceutical Corporation, East Hanover, NJ, USA.

Conflict of interest: L.K., J.J.M., E.J.V., A.P.M., J.K.G., J.M.O.A., M.A.P., and S.D.S. received research support from Novartis Pharmaceuticals; S.Z. is a Novartis Pharmaceuticals employee.


View Abstract